The state of business messaging, 2026

Most published guidance on business messaging describes a world that no longer exists. Four specific changes landed in the last two years, and each one invalidates advice that was correct when it was written.

Meta moved the WhatsApp Business Platform to per-message billing on 1 July 2025, replacing the conversation-based model that had defined the economics since 2022. Utility templates sent inside an open customer service window became free for all businesses at the same time.

What this invalidates: every cost model built on conversation counts, and every piece of advice that treats a 24-hour conversation as a unit you are buying. The rate card and the rules about what is billable have changed more than once; treat the pricing documentation as the source of truth rather than any secondary summary, this one included.

Since February 2025, major US carriers block A2P traffic from unregistered 10DLC numbers outright.

What this invalidates: any launch plan that treats registration as a post-launch task, and any advice describing a degraded-but-functional path for unregistered senders. There is no such path. Campaign vetting can take around ten business days, and a rejection means resubmission.

Apple added RCS in iOS 18, released September 2024. The GSMA finalised Universal Profile 4.0 in March 2026, adding in-thread video calling and richer formatting, with 4.1 following on secure messaging foundations and published work on encryption.

What this invalidates: the standing dismissal of RCS as Android-only. Reach is now a question about operator deployment rather than handset support.

What it does not change: it remains carrier-mediated A2P with registration and content rules, and every deployment still needs an SMS fallback designed as a first-class path.

The EU AI Act's transparency obligations began enforcement on 2 August 2026. People interacting with an AI system must be told, clearly and at first interaction, in a form that cannot easily be overlooked.

What this invalidates: automated support built on the assumption that disclosure is a brand choice. It is now a legal one in the EU, and it stacks with the GDPR rather than replacing any part of it.

Worth stating, because the noise around the above obscures it:

• Delivery rate still measures the network, not your programme.
• Consent is still per channel and per purpose, and a single boolean still cannot express it.
• The inbound direction is still the cheapest and least constrained, and programmes that earn inbound contact still outperform programmes that buy outbound reach.
• Identity across channels is still the hard part, and no vendor has solved it for you.

The noise around four simultaneous changes obscures the durable part, and the durable part is where programmes actually succeed or fail.

Delivery rate still measures the network rather than your programme. Consent is still per channel and per purpose, and it still does not travel between them. The layer above transport — orchestration, identity, evidence — is still where switching is expensive and where every serious difference between vendors lives.

None of the four changes above touches any of that. A team that got the durable part right in 2023 has four adjustments to make. A team that did not has a rebuild, and the four changes are merely what made it visible.

If your messaging programme was designed before mid-2025 and has not been revisited, at least three of its assumptions are now wrong — and the two that will cost you money first are the billing model and the registration requirement.

Start there.

Read together rather than separately, the sequence has a direction: every one of these moves cost, accountability or evidence from the platform onto the sender.

Four changes, one direction: cost, accountability and evidence moved toward the sender. Re-model the economics, move registration before launch, treat RCS as a capability rather than a channel, and disclose automation. Then check that the parts nobody changed are still right, because that is where the expensive problems are.